a. stevanovic, phd, pe - flprite...a. stevanovic, phd, pe. presentation outline •about the project...
TRANSCRIPT
![Page 1: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/1.jpg)
Traffic Demand Patterns and Signal Retiming
Strategies for ITS-Data Rich Arterials
Session 1b - Using ITS to make TSM&O Smarter
2017 Florida Section ITE & ITS FL Annual Meeting
November 2, 2017 – Orlando, FL
A. Stevanovic, PhD, PE
![Page 2: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/2.jpg)
Presentation Outline
• About the project with FDOT
• Research problem – how many signal plans are needed per year?
• Methodology
• Analysis of the results
• What does this really mean? - Discussion
• Conclusions & future plans
![Page 3: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/3.jpg)
• TITLE: Analysis of Traffic Demand Patterns and Signal Retiming Strategies for ITS-data-rich Arterials
• Evaluate robustness of signal timing plans for varying traffic conditions
• Incorporate long-term field data into the decision making process
• Identify reasons for traffic congestion• Develop a microsimulation model and test multiple
representative traffic scenarios• Test existing signal timing plans and develop new ones which
may provide better flows and reduce certain congestion performance measures
• Develop a method which would help FDOT to identify conditions which warrant execution of different timing plans
Project Description
3
![Page 4: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/4.jpg)
Preliminary Results – Analyzing Trends
4
![Page 5: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/5.jpg)
• We know how many TOD signal patterns we need
• How many (off)peak patterns we need per year (DOY)?
• There is no research on this topic
– Difficult to address without having a lot of data
– Methods are not developed/known
– Retiming is costly – making a point that more patterns are needed may not be received well by the community
• This research addresses this problem by:
– Presenting a method to estimate distinctive traffic patterns
– Assigning signal timing plans to those distinctive patterns
– Validating the above assignment through signal optimization
– As a bonus – we get to discuss if we have a case for adaptive system
Focus of this Research Subsection
![Page 6: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/6.jpg)
• Collect field data (mid-block volumes & travel times)
• Store and group data in proper format
• Apply clustering methods
• Identify peak-hour and # of clusters (based on inflection point)
• Decide on data type to use (volumes, travel times or both)
• Identify a representative day for each cluster
• Develop full set of volumes for each representative day (flow-balancing of measured volumes and historical TMCs)
• Populate signal optimization model and optimize signal timings
• Analyze results of signal timing optimization & discuss benefits
• Analyze DOY application patterns
• Make a case for/against Adaptive Traffic Control
Research Method
![Page 7: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/7.jpg)
Sunrise Blvd – A1A to SR7
• One of the busiest corridors in Broward County• Caries around (on some segments) 30,000 veh/day• A segment which caries heavy traffic from another major arterial (US1)• Prone to multiple preemption events (railroad & drawbridge)• Beach access and freeway interchange, a lot of commercial activities
![Page 8: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/8.jpg)
• Turning movement counts (historical and some new)• Travel time – trajectories; probe vehicles with GPS• Travel time – segments; RITIS platform with INRIX & HERE data sources• Speeds – point-based – MVDS • Mid-block volumes – MVDS• Occupancies – MVDS• Split history data – ATMS.now• Preemption events – ATMS.now• Signal timing sheets (basic & coordination)• Signal design plans• Google earth images• AADT volumes – Florida Traffic Online• Video recordings (used for saturation flow rate observations)• Drawbridge activations
• 230-page report just to show the trends from analyzed data
Data – Collection and Handling
# of drawbridge
openings (weekdays in
2016)
![Page 9: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/9.jpg)
• K-means clustering is an unsupervised clustering method whose goal is to partition a given set of observations (data points) (x1, x2, …, xmd) into k clusters by minimizing the within-cluster sum of squares (WCSS):
• where μi is mean point of Si or centroid of cluster i ( i = 1, 2,…, k). For the k-means clustering, number of clusters k need to be specified beforehand. This number of clusters depends on the specific domain application and it can usually be inferred with the help of the “elbow” curve WCSSk = f (k).
Data Clustering – Representative Days
![Page 10: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/10.jpg)
Data Clustering – Where to “draw a line”?
AM Peak hour starts at 8:15 AM
Point where gain of introducing another cluster diminishes
![Page 11: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/11.jpg)
• System inputs
• Distinctive patterns
• Reflect common /group behavior
Cluster Analysis Results – Volumes
![Page 12: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/12.jpg)
Cluster Analysis Results – Travel Times
• System outputs
• Not very distinctive patterns
• Less of causal group behavior
![Page 13: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/13.jpg)
High Traffic Volumes == Long Travel Times?
Travel time –outside circle
Volume –inside circle
![Page 14: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/14.jpg)
Varying of Travel Times for Given Volumes
![Page 15: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/15.jpg)
Volume-based Clusters of AM Peak Traffic
Peak traffic High traffic Moderate traffic Low/Holiday traffic
![Page 16: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/16.jpg)
TT-based Clusters of AM Peak Traffic
Peak traffic High traffic Moderate traffic Low/Holiday traffic
![Page 17: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/17.jpg)
Vol&TT-based Clusters of AM Peak Traffic
Peak traffic High traffic Moderate traffic Low/Holiday traffic
![Page 18: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/18.jpg)
Consistency of Clusters over Multiple Years
![Page 19: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/19.jpg)
Vol&TT-based Clusters of AM Peak TrafficPeak traffic High traffic Moderate traffic Low/Holiday traffic
Do we switch to a separate signal pattern
only for a week?
End of Peak Traffic season?
Beginning of Peak Traffic season?
Distinctive traffic ‘seasons’ or leopard-
fur-like pattern?
Define seasons as groups of days with minimal number of ‘changes’ – days when traffic patterns are different from a
previous day
![Page 20: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/20.jpg)
Identify Seasons for (DOY) Signal Patterns
Many fluctuations –good case for
ATCS?
![Page 21: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/21.jpg)
How many AM Peak patterns and when to implement?
Data for 2015 show two major distinctive signal patterns for AM peak (+ holiday pattern in Dec/Jan) but one of them is effective twice
during the year
Data for 2016 show two major distinctive
signal patterns for AM peak (+ holiday pattern in Dec/Jan)
![Page 22: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/22.jpg)
• Performed for one peak hour (AM)
• Traffic flow balanced based on mid-block volumes
• Historical TMC counts served as foundation for movement flows
• VISTRO – used for its compatibility with VISSIM
• One run performed (to optimize PI) for each set of traffic conditions
• Cross-comparisons were not made (e.g. optimized for low traffic but implemented in high traffic)
• Purpose - various traffic conditions deserve different signal timings – not to estimate benefits of retiming
Signal Optimization
![Page 23: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/23.jpg)
Signal Optimization - Vistro
![Page 24: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/24.jpg)
Balancing of Traffic Flows for Optimization
![Page 25: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/25.jpg)
Benefits of Multiple Signal Patterns (% of PI)
A signal timing plan developed for moderate traffic (recognized by volume data only) generates a PI which is 12.9% lower than a PI generated by an optimal signal timing plan for heavy traffic. This justifies a need for two different plans.
Grey cells indicate cases where the savings/costs are larger than
10%. This is (arbitrarily) chosen as a threshold to warrant
development/implementation of different signal timing patterns.
![Page 26: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/26.jpg)
Conclusions• Traffic data (volumes and travel times) can be used to distinctively recognize traffic
scenarios which may warrant different signal timing plans• Multiple clustering methods revealed that the best results are achieved when both
volumes (inputs) and travel times (outputs) are used to recognize distinctive traffic scenarios
• When signal plans are developed for those scenarios their relevant performance measures were distinctive enough to justify the applied clustering technique
• The compactness of the revealed traffic patterns can help when deciding on what type of traffic signal control should be deployed. – More compact patterns suggest that DOY signal plans are good solutions for weeks/months when
respective patterns were historically observed. – Less compact patterns suggest investigation of the benefits of an adaptive traffic control system. T– In the case study of Sunrise Blvd the results show that use of three additional DOY plans would
properly cover around 70% of the variability in day-to-day traffic.
• Resulting signal timing plans were distinctive enough to yield significantly different performance in majority of the cases (80%).
• Future research – try to generalize findings and find thresholds for some of the developed measures
• The proposed method could be embedded into a decision-support system to suggest periods/days of operations for various signal timing plans
![Page 27: A. Stevanovic, PhD, PE - FLPRITE...A. Stevanovic, PhD, PE. Presentation Outline •About the project with FDOT •Research problem –how many signal plans are needed per year? •Methodology](https://reader036.vdocuments.mx/reader036/viewer/2022071416/6112b3aec684dd5d2e07e4bf/html5/thumbnails/27.jpg)
Acknowledgement & Disclaimer
• Many thanks to Dr. Nikola Mitrovic and FAU GRAs: Djurdjija Mitrovic and Sharmin Chowdhury for their hard work in this study
• Special thanks to FDOT D4 project managers Melissa Ackert and Allison Glunt for their support and invaluable inputs
• Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the Florida Department of Transportation